Derive causal relation .

#1
Hi,

I have a scenario where a relation needs to be established between Medicine and Symptom . Medicine has causal relation with symptom ( Symptom is any side effect) .
The below table has observation which is symptom collected and the last 4 Medication given to the patient .

Every row in table has patient data , where we collect the symptoms and last 4 medication .

We are looking for a way to derive some conclusion if a certain Medicine cause a certain Symptom.

So can we derive if

M3 causes S1.
M4 causes S3 ...etc..

Any pointers towards a right statistical technique is is highly appreciated .

Regards ,
Marc


Untitled.png
 

Attachments

Last edited by a moderator:

hlsmith

Not a robit
#2
I am not sure if this bias has a name, but I call it the white car bias. Look at all car crashes, how many involve a white car. Sixty percent, well white cars cause car accidents. However, you are missing the color of cars not in a car accident. What if white cars represent 80% of cars, it may be that white cars are protective and not a risk factor.

The same scenario is here. I see multiple issues with your project given your brief introduction. There is a reason why clinicians subscribe fervently to correlation does not equal causation. It is easy to procure a spurious associated due to selection bias, information bias, and confounding as well as a lack of being able to layout the causal model prior to touching and analyzing data. Even if you had the full sample of patients on the drugs you will also need to address interactions between drugs and patient differences. The planned scope of this project is very difficult given the scenario you have provided.